Keyboard shortcuts

Press or to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Supported Providers

Caliban is provider-agnostic: you choose which AI provider and model to use at runtime, and the same agent loop, tool engine, and permission system work regardless of which backend answers the requests.

Provider table

Provider--provider valueTransport / accessNotes
AnthropicanthropicDirect HTTPS (api.anthropic.com)Default provider
Anthropic via Bedrock(router only)AWS Bedrock (bedrock-runtime.*)Requires caliban-provider-bedrock; configured via caliban.toml
Anthropic via Vertex(router only)Google Vertex AIRequires caliban-provider-vertex; configured via caliban.toml
OpenAIopenaiDirect HTTPS (api.openai.com/v1)
OpenAI via Azure(router only)Azure OpenAI Serviceazure feature flag on caliban-provider-openai; configured via caliban.toml
GooglegoogleGoogle AI Studio (generativelanguage.googleapis.com)Gemini models
Google via Vertex(router only)Google Vertex AIvertex feature flag; configured via caliban.toml
OllamaollamaLocal HTTP (http://localhost:11434)No API key required

Bedrock, Vertex, and Azure transports are enabled by Cargo feature flags at build time. Binary distributions built by the project team include all features; self-compiled builds must enable the relevant feature (e.g. --features bedrock). These transports can only be selected through the model router — they are not available via the --provider CLI flag.

Capability matrix

ProviderTool useVisionThinkingPrompt caching
AnthropicParallelYesYesExplicit (up to 4 breakpoints)
BedrockParallelYesYesExplicit (mirrors Anthropic)
Vertex (Anthropic)ParallelYesYesExplicit (mirrors Anthropic)
OpenAIParallelYesYes (o-series)Automatic
Azure OpenAIParallelYesYes (o-series)Automatic
Google AI StudioParallelYesNoNone
Google VertexParallelYesNoNone
OllamaBasicModel-dependentModel-dependentNone

Ollama is local

Ollama runs models on your own machine. No API key, no network traffic, no per-token cost. Ideal for fast-classifier routes, offline use, or privacy-sensitive workloads. Capability varies by the specific model you pull.

Multiple providers at once

The model router lets you combine providers: for example, route main-loop turns through Anthropic while using a local Ollama model for fast classification. Each route gets its own provider, model, and resilience policy.